Devstral 2 2512 vs Llama 4 Maverick
Llama 4 Maverick leads Devstral 2 2512 on the Artificial Analysis Intelligence Index, 9.3 to 8.6, and Llama 4 Maverick costs 2.0x less per output token ($1.00 vs $2.00 per million). Updated October 21, 2026; 5 benchmarks have a result for both.
Devstral 2 2512 or Llama 4 Maverick: which to pick
Neither model has a coding score yet, so the two cannot be ranked on coding.
Pick Devstral 2 2512 for a lower cost per solved task ($1.19 against $1.491) and work measured by AA Intelligence Index, where it leads.
- Shared results
- 5
- Devstral 2 2512 only
- 1
- Llama 4 Maverick only
- 80
- Like-for-like categories
- 3 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Neither has a coding score yet.
One side not measured
Intelligence
AA Intelligence Index
Llama 4 Maverick
Llama 4 Maverick leads, 9.3 against 8.6.
Same board, both measured
Agentic work
AA Agentic Index
Devstral 2 2512
Devstral 2 2512 leads, 2.3 against 0.6.
Same board, both measured
Reasoning
GPQA Diamond
No call
Only Llama 4 Maverick has a GPQA Diamond result, 69.4%.
Directional only
Math
AIME
No call
Only Llama 4 Maverick has a AIME result, 25.2%.
Directional only
Knowledge
MMLU Pro
No call
Only Llama 4 Maverick has a MMLU Pro result, 79.4%.
Directional only
Web design
Design Arena: websites
No call
Only Llama 4 Maverick has a Design Arena: websites result, 871.
Directional only
Long documents
Listed context window
Llama 4 Maverick
Llama 4 Maverick takes 524K tokens against 262K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Llama 4 Maverick
Devstral 2 2512 $0.0014, Llama 4 Maverick $0.0008 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
Devstral 2 2512
Devstral 2 2512 $1.19, Llama 4 Maverick $1.491.
Published attempt cost and pass rate
At a glance
- Coding index
- Devstral 2 2512 not benchmarked, Llama 4 Maverick not benchmarked.
- Widest category
- Coding, 17.4 points apart on average over 3 shared benchmarks.
- Widest single gap
- SWE-bench Verified: Devstral 2 2512 53.8% against Llama 4 Maverick 21.0%, 32.8 points apart on a 0 to 100 scale.
- Output price
- Devstral 2 2512 $2.00 against Llama 4 Maverick $1.00 per million output tokens, 2.0x apart.
- Context window
- Devstral 2 2512 takes 262K tokens, Llama 4 Maverick 524K.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | Devstral 2 2512 | Bars | Llama 4 Maverick | Ties |
|---|---|---|---|---|
| Coding | 3 | 0 | ||
| Intelligence | 0 | 1 | ||
| Agentic work | 1 | 0 | ||
| All shared boards | 4 | 1 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index30
| Benchmark | Devstral 2 2512 | Llama 4 Maverick | Leader |
|---|---|---|---|
| Aider polyglot | Not measured | 15.6% | Only Llama 4 Maverick measured |
| SWE-bench Verified | 53.8% | 21.0% | Devstral 2 2512 |
| AA Coding Index | 31.3 | 16.3 | Devstral 2 2512 |
| LiveCodeBench (Vals AI) | 51.8% | 47.3% | Devstral 2 2512 |
| SWE-bench Verified (Vals AI) | 62.8% | Not measured | Only Devstral 2 2512 measured |
| LiveCodeBench (Artificial Analysis run) | Not measured | 39.7% | Only Llama 4 Maverick measured |
| SciCode | Not measured | 31.7% | Only Llama 4 Maverick measured |
| Terminal-Bench Hard (Artificial Analysis run) | Not measured | 6.8% | Only Llama 4 Maverick measured |
| WeirdML | Not measured | 24.5% | Only Llama 4 Maverick measured |
| ALE-Bench | Not measured | 173.0 | Only Llama 4 Maverick measured |
| LMArena Text: coding | Not measured | 1301 | Only Llama 4 Maverick measured |
| LMArena Text: software and IT | Not measured | 1299 | Only Llama 4 Maverick measured |
| Neither model has a result on the other 27 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Devstral 2 2512 | Llama 4 Maverick | Leader |
|---|---|---|---|
| AA Intelligence Index | 8.6 | 9.3 | Llama 4 Maverick |
| Epoch Capabilities Index | Not measured | 132.2 | Only Llama 4 Maverick measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | Devstral 2 2512 | Llama 4 Maverick | Leader |
|---|---|---|---|
| AA Agentic Index | 2.3 | 0.6 | Devstral 2 2512 |
| τ²-Bench Telecom | Not measured | 17.8% | Only Llama 4 Maverick measured |
| Neither model has a result on the other 39 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index13 boards, none shared
| Benchmark | Devstral 2 2512 | Llama 4 Maverick | Leader |
|---|---|---|---|
| GPQA Diamond | Not measured | 69.4% | Only Llama 4 Maverick measured |
| MMMU Pro (Vals AI) | Not measured | 71.7% | Only Llama 4 Maverick measured |
| GPQA Diamond (Artificial Analysis run) | Not measured | 67.1% | Only Llama 4 Maverick measured |
| Humanity's Last Exam (Artificial Analysis run) | Not measured | 4.9% | Only Llama 4 Maverick measured |
| ARC-AGI | Not measured | 4.4% | Only Llama 4 Maverick measured |
| ARC-AGI-2 | Not measured | 0.0% | Only Llama 4 Maverick measured |
| SimpleBench | Not measured | 27.7% | Only Llama 4 Maverick measured |
| LMArena Text: hard prompts | Not measured | 1281 | Only Llama 4 Maverick measured |
| LMArena Text: expert | Not measured | 1259 | Only Llama 4 Maverick measured |
| LMArena Text: hard prompts in English | Not measured | 1295 | Only Llama 4 Maverick measured |
| LMArena Vision: diagrams | Not measured | 1134 | Only Llama 4 Maverick measured |
| LMArena Vision: homework | Not measured | 1164 | Only Llama 4 Maverick measured |
| Kagi LLM Benchmark | Not measured | 55.9% | Only Llama 4 Maverick measured |
| Neither model has a result on the other 9 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index7 boards, none shared
| Benchmark | Devstral 2 2512 | Llama 4 Maverick | Leader |
|---|---|---|---|
| AIME | Not measured | 25.2% | Only Llama 4 Maverick measured |
| MATH 500 (Vals AI) | Not measured | 85.2% | Only Llama 4 Maverick measured |
| MGSM (Vals AI) | Not measured | 92.4% | Only Llama 4 Maverick measured |
| AA Math Index | Not measured | 19.3 | Only Llama 4 Maverick measured |
| AIME 2025 (Artificial Analysis run) | Not measured | 19.3% | Only Llama 4 Maverick measured |
| LMArena Text: math | Not measured | 1299 | Only Llama 4 Maverick measured |
| LMArena Text: mathematical work | Not measured | 1295 | Only Llama 4 Maverick measured |
| Neither model has a result on the other 6 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index13 boards, none shared
| Benchmark | Devstral 2 2512 | Llama 4 Maverick | Leader |
|---|---|---|---|
| MMLU Pro | Not measured | 79.4% | Only Llama 4 Maverick measured |
| LegalBench (Vals AI) | Not measured | 77.8% | Only Llama 4 Maverick measured |
| CorpFin (Vals AI) | Not measured | 49.7% | Only Llama 4 Maverick measured |
| TaxEval (Vals AI) | Not measured | 66.6% | Only Llama 4 Maverick measured |
| MortgageTax (Vals AI) | Not measured | 58.5% | Only Llama 4 Maverick measured |
| MedQA (Vals AI) | Not measured | 43.3% | Only Llama 4 Maverick measured |
| MedCode (Vals AI) | Not measured | 36.5% | Only Llama 4 Maverick measured |
| MMLU Pro (Artificial Analysis run) | Not measured | 80.9% | Only Llama 4 Maverick measured |
| GeoBench | Not measured | 52.0% | Only Llama 4 Maverick measured |
| LMArena Text: science | Not measured | 1288 | Only Llama 4 Maverick measured |
| LMArena Text: medicine and healthcare | Not measured | 1275 | Only Llama 4 Maverick measured |
| LMArena Text: legal and government | Not measured | 1286 | Only Llama 4 Maverick measured |
| LMArena Text: business and finance | Not measured | 1260 | Only Llama 4 Maverick measured |
| Neither model has a result on the other 5 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
| Benchmark | Devstral 2 2512 | Llama 4 Maverick | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 871 | Only Llama 4 Maverick measured |
| Design Arena: UI components | Not measured | 899 | Only Llama 4 Maverick measured |
| Design Arena: game dev | Not measured | 838 | Only Llama 4 Maverick measured |
| Design Arena: data viz | Not measured | 882 | Only Llama 4 Maverick measured |
| Design Arena: 3D | Not measured | 901 | Only Llama 4 Maverick measured |
| Design Arena: all code categories | Not measured | 880 | Only Llama 4 Maverick measured |
| Neither model has a result on the other 3 boards here | |||
GeneralDirectional onlyPublic boards, shown as published, kept out of the index31 boards, none shared
| Benchmark | Devstral 2 2512 | Llama 4 Maverick | Leader |
|---|---|---|---|
| SAGE (Vals AI) | Not measured | 35.7% | Only Llama 4 Maverick measured |
| MedScribe (Vals AI) | Not measured | 54.2% | Only Llama 4 Maverick measured |
| IFBench | Not measured | 43.0% | Only Llama 4 Maverick measured |
| Long context reasoning (AA-LCR) | Not measured | 50.0% | Only Llama 4 Maverick measured |
| DTBench | Not measured | 61.9% | Only Llama 4 Maverick measured |
| LMCA | Not measured | 15.9% | Only Llama 4 Maverick measured |
| Fiction.LiveBench | Not measured | 46.2% | Only Llama 4 Maverick measured |
| Creative writing (Lech Mazur) | Not measured | 62.0% | Only Llama 4 Maverick measured |
| LMArena Text | Not measured | 1287 | Only Llama 4 Maverick measured |
| LMArena Text: instruction following | Not measured | 1267 | Only Llama 4 Maverick measured |
| LMArena Text: creative writing | Not measured | 1266 | Only Llama 4 Maverick measured |
| LMArena Vision | Not measured | 1142 | Only Llama 4 Maverick measured |
| LMArena Text: English | Not measured | 1303 | Only Llama 4 Maverick measured |
| LMArena Text: non-English | Not measured | 1269 | Only Llama 4 Maverick measured |
| LMArena Text: Chinese | Not measured | 1277 | Only Llama 4 Maverick measured |
| LMArena Text: German | Not measured | 1291 | Only Llama 4 Maverick measured |
| LMArena Text: Japanese | Not measured | 1206 | Only Llama 4 Maverick measured |
| LMArena Text: Korean | Not measured | 1203 | Only Llama 4 Maverick measured |
| LMArena Text: French | Not measured | 1259 | Only Llama 4 Maverick measured |
| LMArena Text: Spanish | Not measured | 1293 | Only Llama 4 Maverick measured |
| LMArena Text: Russian | Not measured | 1286 | Only Llama 4 Maverick measured |
| LMArena Text: Polish | Not measured | 1275 | Only Llama 4 Maverick measured |
| LMArena Text: multi-turn | Not measured | 1289 | Only Llama 4 Maverick measured |
| LMArena Text: longer queries | Not measured | 1280 | Only Llama 4 Maverick measured |
| LMArena Text: writing and language | Not measured | 1277 | Only Llama 4 Maverick measured |
| LMArena Text: entertainment and media | Not measured | 1250 | Only Llama 4 Maverick measured |
| LMArena Vision: English | Not measured | 1150 | Only Llama 4 Maverick measured |
| LMArena Vision: Chinese | Not measured | 1141 | Only Llama 4 Maverick measured |
| LMArena Vision: reading text | Not measured | 1143 | Only Llama 4 Maverick measured |
| LMArena Vision: humor | Not measured | 1100 | Only Llama 4 Maverick measured |
| LMArena Vision: creative writing | Not measured | 1099 | Only Llama 4 Maverick measured |
| Neither model has a result on the other 15 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Devstral 2 2512 | Llama 4 Maverick | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
Llama 4 Maverick costs less
Repository review
50K input + 3K output tokens
Llama 4 Maverick costs less
Agent loop
200K input + 10K output tokens
Llama 4 Maverick costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | Mistral Devstral 2 2512 | Meta Llama 4 Maverick |
|---|---|---|
| Maker | Mistral AI | Meta Llama |
| Weights | Open weights | Open weights |
| Released | December 9, 2025 | April 5, 2025 |
| Context window | 262K | 524K |
| Input $/Mtok | $0.40 | $0.35 |
| Output $/Mtok | $2.00 | $1.00 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 1 | 0 |
| Cost per solved task | $1.19 | $1.491 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Devstral 2 2512 better than Llama 4 Maverick for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. Pick Devstral 2 2512 for a lower cost per solved task ($1.19 against $1.491) and work measured by AA Intelligence Index, where it leads.
Which is cheaper, Devstral 2 2512 or Llama 4 Maverick?
Devstral 2 2512 costs $0.40 per million input tokens and $2.00 per million output tokens; Llama 4 Maverick costs $0.35 and $1.00. Per solved SWE-bench Verified task, Devstral 2 2512 costs $1.19 and Llama 4 Maverick $1.491.
Which has the bigger context window, Devstral 2 2512 or Llama 4 Maverick?
Llama 4 Maverick, with 524K tokens against 262K.
How many benchmarks is the Devstral 2 2512 vs Llama 4 Maverick comparison based on?
5 benchmarks have a result for both models as of October 21, 2026. Devstral 2 2512's coding index rests on 1 independent benchmark and Llama 4 Maverick's on 0. Where only one model has a result, the page says so instead of naming a winner.